Search for a command to run...
Cybersecurity in digital finance has become a critical area due to the growing number of online transactions and the increasing sophistication of financial fraud. Digital platforms, mobile wallets, and real-time payment systems are widely adopted, but they often face threats such as identity theft, unauthorized access, transaction manipulation, and synthetic fraud. Traditional rule-based fraud detection methods fall short in identifying complex and adaptive attack patterns, especially in high-speed environments. This paper introduces an artificial intelligence (AI)powered framework for fraud detection and risk management tailored to digital finance systems. The proposed model brings together supervised learning, anomaly detection, and behavioral analytics. A deep learning model processes transaction sequences, user metadata, and device information to capture both shortterm anomalies and long-term behavioral shifts. The approach supports real-time scoring and risk classification by integrating autoencoder-based anomaly filters and ensemble classifiers. A risk-aware decision module also adjusts fraud detection thresholds based on transaction value and context. To evaluate the proposed framework, a real-world financial transaction dataset with labeled fraud instances was used. The model achieved high performance across multiple metrics, including an F1-score of 0.94, an AUC of 0.97, and a precision of 92.8 %. Compared to baseline models such as logistic regression, decision trees, and standard neural networks, the proposed framework showed significant improvements in early fraud detection and low false positive rates. These results highlight its ability to reduce financial loss, maintain transaction security, and support trust in digital payment systems. The approach is scalable to large datasets and adaptable to new fraud trends, making it suitable for modern financial infrastructures.